DEPARTMENT OF PHYSICS DISSERTATION DEFENSE: Olivia Ghosh

Public zoom link:  https://stanford.zoom.us/j/94422946553?pwd=nyVkEyDE3yb2yMkxlFLAfjhUrFTL8y.1

Password: Email physicsstudentservices [at] stanford.edu (physicsstudentservices[at]stanford[dot]edu) for password

Title: 

Discovering structure in fitness maps across space and time

Abstract:

 Evolutionary forces govern the generation, maintenance, and loss of the extensive biological diversity on Earth. The key function that governs the course of adaptive evolution is fitness, which depends both on an organism’s genotype and the environment, and determines which individuals will be successful in the competition for survival. Understanding the mapping between genotypes, phenotypes, and fitness is crucial if we want to make predictions about how evolution will unfold, but it is a major challenge because this mapping is ostensibly high-dimensional – there are billions of potential genetic mutations, each with an unknown set of phenotypic effects, and each with unknown consequences in a given environment. In the first part of this thesis, we address this problem by building a bottom-up model of fitness in a spatially varying environment in order to understand the impact of spatial structure on the evolution of human gut bacteria. In the second part, we take a top-down approach by leveraging fitness variation due to environmental perturbations – previously assumed only to complicate fitness landscapes and hinder prediction – to infer underlying structure in genotype-phenotype-fitness maps. By measuring the fitness effects of thousands of lab-adaptive mutations in Saccharomyces cerevisiae across a range of both subtle and more pronounced perturbations, we use patterns in fitness covariation to understand both how many phenotypes matter to fitness, and how these phenotypes shift as the environment changes. We then use the same top-down approach to interrogate a genetically and ecologically diverse collection of S. cerevisiae strains that were isolated from nature. We apply environmental perturbations and measure fitness to characterize the functional consequences of natural genetic variation, and we experimentally evolve the pool of natural strains to explore the longer-term outcomes of competition within species. In the cases of both the adaptive mutants and the natural isolates, we find that low-dimensional, linear models capture the empirical fitness landscapes remarkably well, suggesting that the structure underlying evolution is often simple enough to be learned from data and leveraged for prediction. More broadly, we show that top-down approaches are useful tools for contending with the high-dimensional challenges posed by evolutionary questions.